The father of ChatGPT becomes a balance master: I want both money and AI technology security

News on April 3, OpenAI’s ultimate goal is to develop safe general artificial intelligence. The company’s CEO, Sam Altman, needs to both use technology revenue to drive continuous advancement of artificial intelligence and use non-profits to ensure the safety of artificial intelligence technology. His close friends say it's the art of balance.

Altman, the 37-year-old CEO of OpenAI, has always dreamed that computers could communicate with and learn from each other like humans. He was brought to the forefront of the artificial intelligence craze with ChatGPT. Altman said that what he remembers most clearly is staying up late as a child in his bedroom playing the Macintosh LC II, which he got as an eighth birthday present. At that time, Altman suddenly realized: "One day, computers will learn to think."

In recent months, Altman has not only been building the future of artificial intelligence, but also working hard to its commercialization. In November last year, the OpenAI company he led released ChatGPT, a chatbot that can write articles like humans and has become one of the most viral products in the history of technology. Investor documents show that as ChatGPT became popular across the Internet, OpenAI grew from a small non-profit organization to a multi-billion dollar company at a near record speed. This is partly due to OpenAI establishing a for-profit unit and receiving $13 billion in financing from Microsoft.

Such success is also part of a delicate balancing act. Altman said he also worries about what will happen if artificial intelligence is introduced into society with reckless abandon. He and others co-founded OpenAI, a nonprofit research organization, eight years ago when Altman argued that using profits to drive the development of powerful models for artificial intelligence was dangerous.

Altman is wary of using profits to incentivize AI development. He said he does not directly hold shares in OpenAI, which is rare in Silicon Valley. You know, the founders of successful Silicon Valley startups usually get rich through the equity they hold.

Altman said of the economic benefits: “Like most people, I like to watch the wealth numbers go up, but I just don’t want that to be the full factor.” OpenAI said Altman The salary is "moderate", but no specific figure was disclosed. Altman revealed that he holds a small stake in a venture fund investing in OpenAI, but that it is "inconsequential."

Altman made a lot of money in his early years by investing in start-ups, "more than I needed." He owns three properties, including a mansion in a San Francisco neighborhood and a vacation home in Napa Valley. He hired dozens of people to manage properties and family investments.

Altman’s goal is to establish a new order and use machines to liberate people so that they can freely engage in more creative work. In Altman’s scenario, everyone would receive a basic income to compensate for the jobs being replaced by artificial intelligence. He even believes that humans will be very fond of artificial intelligence, and advanced chatbots can become "an extension of people's will."

Altman said that in the long term, he hopes to establish a global governance structure to oversee decisions about future artificial intelligence and gradually reduce the influence of OpenAI’s executive team on artificial intelligence technology.

Supporters say Altman is the ideal candidate to lead OpenAI. But many people say that Altman's business acumen is too developed and he is always immersed in Silicon Valley thinking, unable to bear the technological revolution that is reshaping business and social life.

Investor documents show that OpenAI and Microsoft signed an agreement worth $10 billion in January this year, and Microsoft will own 49% of OpenAI’s for-profit unit. This corporate partnership, coupled with Altman's aggressive push to commercialize artificial intelligence technology, frustrated OpenAI's early leaders. They argue that these decisions violate initial promises to develop AI outside of shareholder influence.

Tesla CEO Elon Musk also criticized OpenAI. Musk participated in the founding of OpenAI in 2015, but left in 2018 due to disputes over control and development direction. In February this year, Musk said on Twitter that OpenAI was originally established as an open source non-profit organization "to check and balance Google, but now it has become a company with closed resources, pursuing maximum profits, and is actually controlled by Microsoft." company. That was not my intention at all." Altman paused when asked about Musk's criticism. "I like Musk," he finally replied, "and I pay attention to what he says." In an open letter released last week, including Musk, Apple co-founder Steve Wozniak Silicon Valley celebrities including Steve Wozniak have called on the entire industry to suspend the development of artificial intelligence more advanced than OpenAI's latest technology GPT-4 within six months to avoid a situation where the technology competition may get out of control.

People have long imagined general artificial intelligence that is not just good at generating text or images, but can understand and learn as well as humans, or even perform better than humans. Altman said that OpenAI does not necessarily need to be the first team to develop general artificial intelligence, but their ultimate mission is to develop safe general artificial intelligence.

According to investor documents, OpenAI has set a profit cap for investors, and any income exceeding a certain level will flow to the non-profit organization to which it belongs. The specific value of the profit ceiling ranges from 7 times to 100 times the invested capital, depending on the investment time. OpenAI and Microsoft have also formed a joint security committee that includes Altman and Microsoft Chief Technology Officer Kevin Scott. If the committee believes that the products released by Microsoft and OpenAI are too dangerous, it has the power to withdraw them.

Billionaire venture capitalist Peter Thiel, an early donor to OpenAI, is a close friend of Altman. He said the possibility of general artificial intelligence gave Altman the idea that some similar technology created the universe. Thiel has long supported the idea that humans and machines will one day merge with each other.

After the release of ChatGPT, a series of competing artificial intelligence systems also emerged. In February this year, Google announced that it was testing its own chatbot Bard and would open it to the public in March.

"They raced to issue press releases," Altman said of his competitors. "Obviously, they're behind now."

OpenAI committed in its founding charter to abandon existing research efforts if another project came closer to developing general artificial intelligence than it did. The company said this was to put human safety first. , avoiding the race to build dangerous artificial intelligence systems.

OpenAI’s headquarters in San Francisco’s Mission District has offices filled with stone fountains, succulents and ferns. Altman gushes about the central staircase he designed , so that the company's 400 employees have the opportunity to meet up close every day. From Monday to Wednesday, employees must also work face to face. The building also has a university-style cafeteria, self-service bar and library. The design of the library is Ault A combination of Mann's favorite Parisian bookstore and the design inspiration of the Bender Room, a quiet study space on the top floor of the Stanford University Library.

Altman wore the typical technology CEO outfit: a gray sweatshirt, jeans and a shiny white shirt Sneakers. He said he grew up very modestly.

Altman grew up in the suburbs of St. Louis, the eldest of four children. His mother was a dermatologist and his father worked as a lawyer. Various jobs. Altman's father's real career, family said, was running an affordable housing nonprofit and spent years trying to revitalize downtown St. Louis.

Altman said , one lesson his father taught him was, “You should try your best to help others. Even if you don’t have time, you have to find a way. "

Altman's mother said that her son started using the family's video recorder when he was 2 years old, and changed his own ticket home from summer camp when he was 13. She said that when Altman was studying In third grade, he started helping local public school teachers with computer problems. In middle school, he transferred to the private John Burroughs School.

“The teachers liked him because he was really, really Smart and hard-working, but also very social," said then-principal Andy Abbott. "He was fun and had a great personality. ”

Altman later went to Stanford University to do research in an artificial intelligence lab. As a sophomore, he co-founded Loopt, a location-based social network service company, which later became the startup accelerator Y Part of Combinator's first class. Altman later left the school.

Loopt never took off, selling in 2012 for $434 million, close to funding from investors including Sequoia Capital Um.

Subsequently, Altman founded a venture fund and made friends such as Thiel and Y Combinator co-founder Paul Graham. These people helped Altman Mann's company built itself into an influential brokerage in Silicon Valley. Altman has used his own money to successfully invest in dozens of companies, including cloud software company Asana and message board site Reddit.

In and During his time at Y Combinator, Altman became increasingly concerned that large research labs like DeepMind were developing potentially dangerous AI technology out of the public eye. Musk was equally worried about powerful AI machines taking over human society.

Both Altman and Musk decided it was time to build their own labs. Eventually, a team they were part of pledged $1 billion in founding funding to the nonprofit OpenAI.

OpenAI faltered at the beginning, trying some small projects, such as teaching robots how to solve Rubik's cubes. In 2017, OpenAI laid off a large number of employees.

OpenAI researchers quickly concluded that, The most likely way to achieve general artificial intelligence lies in large language models, which are computer programs that imitate the way humans read and write. Altman said that such models need to be trained on large amounts of text content and require a lot of computing power. But as a company As a non-profit organization, OpenAI does not have the conditions for large-scale financing.

"We didn't realize how expensive this project was going to be," Altman said. "We didn't know it at the time."

Altman said that year he explored various There are several options to raise more funding for OpenAI, such as seeking government funding or launching a new cryptocurrency. He said: "No one was willing to fund the research, and it was a really difficult time."

According to people familiar with the matter, tensions between Altman and Musk continued to increase. Musk was frustrated with the slow progress of research and wanted to gain more control over OpenAI.

OpenAI executives eventually came up with an entirely different idea, creating a for-profit division, OpenAI LP, that would report to a non-profit parent organization.

Reid Hoffman, co-founder of recruitment platform LinkedIn, was an advisor to OpenAI at the time and later became a member of the company’s board of directors. He said the idea is to attract investors who are eager to benefit from the commercialization of OpenAI technology and accelerate the development of OpenAI. "You want to be the first to achieve something, you want to set the standard," Hoffman said. "That's why speed of development becomes part of the moral and ethical issue."

People familiar with the matter said that this decision further alienated Musk’s operating philosophy from OpenAI. In February 2018, he parted ways with OpenAI.

Former employees who attended the meeting said Musk announced his departure at an all-staff meeting. Musk explained at the time that he thought he would have a better chance of developing general artificial intelligence through Tesla, where he would have access to more resources.

Former employees say a young researcher questioned whether Musk had considered safety issues at the time. They said Musk became visibly frustrated and called the intern a "moron," leaving employees stunned. It was the last time many of them saw Musk in person.

Soon after, an OpenAI executive commissioned a “Fool” trophy for the young researcher and later gave it to him on a pillow. "You've got to have some fun," Altman said. "This is part of the company culture."

Musk's departure became a turning point for OpenAI. Later that year, OpenAI leadership informed employees that Altman would lead the company. He officially became OpenAI CEO and helped create the for-profit unit in early 2019.

Of the $1 billion originally pledged in donations, OpenAI said it received approximately $130 million. With the for-profit arm established, OpenAI no longer needs further donations. Musk said on Twitter that he donated approximately $100 million to OpenAI.

According to former employees who attended the meeting, Musk announced his resignation at an all-staff meeting. Musk explained at the time that he believed that through Tesla, he would have access to more resources and thus have a better chance of developing general artificial intelligence.

Former employees said that a young researcher once asked Musk if he had considered safety issues. They said Musk was visibly frustrated and shocked employees by calling the intern a "sucker." It was the last time many of them saw Musk in person.

Soon after, OpenAI executives commissioned a “Fool” trophy and placed it on the researcher’s pillow as a gift. Altman said: "You have to have some fun, and that's part of the company culture."

Musk's departure became a turning point for OpenAI. Later that year, OpenAI's leadership informed employees that Altman would lead the company. He officially became OpenAI's CEO and began creating the for-profit unit in 2019.

Of the initially announced $1 billion donation, OpenAI said it has received approximately $130 million. After creating a for-profit unit, OpenAI no longer needs further donations. Musk said on Twitter that he donated about $100 million to OpenAI.

Meanwhile, Altman began looking for investors. In the summer of 2018, he was attending the Allen & Co. annual meeting in Sun Valley, Idaho, and happened to run into Microsoft CEO Satya Nadella in a stairwell. Altman introduced OpenAI to Nadella and got his interest. That winter, formal talks began.

"I remember coming back later and saying, this is the only right partner," Altman recalled. "They got something safe, they got artificial intelligence. And they have capital and a lot of computing power."

Former employees of OpenAI said that Altman promptly announced the contents of the contract during the negotiation process. And put a full staff on the job to allay concerns that the partnership would be seen as contradicting OpenAI's original promise to develop artificial intelligence outside the enterprise world.

However, some employees still believe this is a deal with the devil. Dario Amodei, OpenAI's chief security researcher, and his deputies worried that the deal would allow Microsoft to sell OpenAI technology before it has passed adequate security testing. They believe that OpenAI technology is still far from ready for a large-scale release, let alone working with one of the largest software companies in the world. They worry that technology could malfunction or be misused, causing unpredictable harm.

Some former employees said Amodei was also concerned that the deal would tie OpenAI to Microsoft and make it harder for OpenAI to fulfill its founding charter promise to achieve artificial general intelligence if it were first achieved. Will turn around and help another project.

Microsoft initially invested $1 billion in OpenAI. The deal brings needed funding to OpenAI, but also comes with some limitations. OpenAI agreed to use Microsoft giant servers to train artificial intelligence models only through Azure cloud services, and licensed OpenAI technology to Microsoft products for use.

Nadella said in a recent interview: "When you stand on the edge of a cliff, the only thing you can do is jump off and hope that you can land stably." He called this a process of platform transformation.

Gary Marcus, professor emeritus of psychology and neuroscience at New York University and co-founder of a machine learning company, said: “This deal completely undermines OpenAI’s efforts to obtain nonprofit status. The principles followed.”

Marcus believes that Altman “pulled a 180-degree turn and, in my opinion, was just paying lip service to concern for human safety.”

Altman disagreed. “What’s extraordinary about Microsoft as a partner is that it allows us to retain all the principles that we think are important to the mission,” he said, including investment. A cap on earnings and a promise to turn around and help another project if general artificial intelligence is achieved first, and so on.

This funding will undoubtedly accelerate OpenAI’s research and development progress, allowing researchers to obtain the critical computing power needed to improve large language models. OpenAI quickly developed a more powerful language model, GPT-3, and sold the technology to developers via an application programming interface (API) in June 2020.

According to former employees, Altman and Amodei clashed again over API release issues. Amodei believes that product releases should be limited to smaller user groups and conducted in stages to reduce public attention and give security teams more time to test.

A few months later, Amodei left OpenAI with others to start a competing artificial intelligence lab, Anthropic. Altman said: "They disagree with us on how to get safe general artificial intelligence."

Anthropic has received more than $300 million in investment from Google and released artificial intelligence in March this year. Intelligent chat robot Claude, developers can call this robot through API.

According to a recent investor report, Anthropic stated that it is “committed to large-scale commercialization” to develop safe general artificial intelligence, and plans to release a commercialization method in September that is “in line with the commitment”. Anthropic was founded as an artificial intelligence security and research company with plans to potentially create business value from the product.

Investor documents show that Microsoft invested a total of $3 billion in OpenAI in the three years after the original deal.

In November last year, ChatGPT gained more than 1 million registered users in just five days after its release. This speed even surprised Altman. Prior to this, OpenAI launched DALL-E 2, which can generate complex images based on text prompts.

After ChatGPT spread, people used this chatbot for various interesting applications, such as writing poems or planning birthday parties, which aroused heated discussions on the Internet.

According to UBS analysts, by February this year, the number of active users of ChatGPT had reached 100 million, which is the fastest popularization of consumer applications in history.

Altman’s close associates praise his ability to balance OpenAI’s priorities. Thiel said no one is better suited to strike a balance between "idealism" and "reality."

Altman said he delayed the release of the latest version of GPT-4 until March this year in order to conduct more security testing. Many users have reported some problems when using the GPT-4 model. For example, when integrated into Microsoft's Bing search, GPT-4 will make up some wrong answers.

Altman said: "The right approach is to get people involved, explore these systems, study them and learn how to make them more secure."

According to the document, in After recouping the initial investment, Microsoft will receive about 49% of the profit before OpenAI reaches the revenue ceiling, which is higher than the 21% previously arranged. Non-profit parent company OpenAI will receive residual proceeds.

In addition to OpenAI, Altman has invested in other projects, including Worldcoin, a company he co-founded, with the goal of providing cryptocurrency services to everyone.

In recent years, Altman has invested almost all of his working capital in two companies. Altman said his $375 million investment in Helion Energy, which is researching how to create carbon-free energy from nuclear fusion, is close to "creating a net energy gain in a real demonstration."

In addition, he invested US$180 million in Retro, with the goal of extending human life by 10 years through innovative technologies such as "cell reprogramming, plasma therapy and autophagy."

Altman said these issues are much easier to deal with ethically than with artificial intelligence. "If you're doing nuclear fusion, it's definitely a benefit. It's a good thing," he said. "If you're developing artificial intelligence, it could be very good or it could be very bad."

The above is the detailed content of The father of ChatGPT becomes a balance master: I want both money and AI technology security. For more information, please follow other related articles on the PHP Chinese website!